The tech world is once again caught up in a shiny new idea: that artificial intelligence, especially generative AI, will revolutionize how fast software gets built. For decades, similar promises have emerged—from offshoring in the 2000s to microservices in the 2010s—each claiming to be the magic solution for making developers more productive. Today, the claim is that AI can instantly write code, speeding up delivery and reducing costs. But is that really happening?

Understanding the Real Bottlenecks in Software Development

While it’s tempting to think that AI speeds up the act of writing code itself, the truth is that coding is rarely the slowest part of software projects. The real challenge lies elsewhere: deciding what to build, planning the approach, integrating new features into existing systems, ensuring security and compliance, and then maintaining and operating the final product. AI can assist with syntax, boilerplate, and repetitive tasks, but it doesn’t address these complex, human-driven steps.

What’s more, AI makes complexity cheaper and easier to create, which can be a double-edged sword. It lowers the barrier to building more features or systems, but it also risks increasing chaos if developers aren’t given proper guidance and structure. Without proper frameworks or platforms, the productivity gains from AI can quickly be overshadowed by disorganization and technical debt.

The Evidence Is Mixed—And That’s No Accident

Research and real-world experiments show that AI’s impact on developer productivity isn’t straightforward. For example, a controlled study found that experienced open source developers took about 19% longer to complete tasks when using AI tools, even though they believed they would be faster. On the other hand, GitHub reported that developers could complete specific tasks faster with AI-assisted tools like Copilot and felt better about their experience.

This inconsistency highlights an important point: AI’s effectiveness depends heavily on the environment in which it is used. When integrated into a well-structured system with clear processes and guardrails, AI can indeed accelerate development. But in fragmented, chaotic environments, it can amplify confusion and inefficiency.

The Systems Perspective: Environment Matters More Than the Tool

AI isn’t just a tool; it’s part of a larger system. Its success relies on how well organizations set up their workflows, standards, and platforms to support AI-driven development. Long before prompt engineering became a buzzword, experts argued that giving developers too much freedom without proper boundaries leads to sprawl and integration headaches.

Generative AI accelerates these issues because it removes friction that might have slowed poor decision-making. Leaders often confuse shipping more code with being truly productive. If productivity is defined as just output, AI seems like a miracle. But in reality, code is an asset only if it’s secure, maintainable, and aligned with business goals. Without that, more code can become a liability—adding complexity without value.

Ultimately, AI’s role is to support better systems, not just faster coding. Without the right environment, AI can make things worse, creating more chaos instead of progress. The key is building platforms and guardrails that channel AI’s power into meaningful, manageable growth.

Inspired by

• https://www.infoworld.com/article/4125409/is-ai-killing-productivity.html

Sources

• metr.org
• github.blog
• dora.dev
• microsoft.com
• pragmaticengineer.com